Mapping the Landscape of Research on Chinese Social Media: What We Know and What We Don’t Know

Jonathan J. H. Zhu with T. Q. Peng, H. Liang, J. Qin, Z. Z. Wang, H. X. Chen, & B. L. Li

Research Seminar, October 6, 2014 Dept. of Media & Communication, CityU

“The Mapping Project”

Mapping the landscape of Internet studies (Peng et al., 2013, New Media & Society)

Mapping the intellectual structure of e-Health research (Jiang et al., 2015, International Journal of Medical Informatics)

Mapping the landscape of research on Chinese social media

2

3

Source: http://www.techinasia.com/china-social-media-landscape-infographic-2014/

Mapping Landscape of Chinese Social Media

• IM • Vdeo/Music • Blog • Microblog • SNS • BBS • Mobile Social • Social Life • E-Commerce

4

Mapping Research on Chinese Social Media?

We need an organizational framework!

The 5W Model for Assessment

Harold Lasswell (1948): Communication is a

process of who says what to whom through which

channel with what effects.

5

Communicator Content Channel Audience Effects

6

SCI/SSCI Publications on Chinese Social Media (N=138)

0%

20%

40%

60%

2009 2010 2011 2012 2013 2014

a. Annual Growth

SCI (103)

SSCI (35)

d. Leading Institutions Rank SCI SSCI

1 CAS Nanyang 2 Tsinghua Tsinghua 3 Peking Peking 3 Beihang Beihang 3 BUPT CityU HK 3 SJTU HUST

26% 46%

12% 14% 14%

30%

44% 14%

0%20%40%60%80%

100%

SCI SSCI

b. Leading Disciplines

Computer

Engineering

Business

Psych

Communication

Others

c. Leading Nationals

SCI SSCI

Distribution across 5Ws

Communicator 16%

Content 15% Channel

5%

Audience 29%

Effects 16%

System Design 19%

For Communicator

4%

For Audience 7%

For Content 6%

Research Method 17%

7

Differences between SCI and SSCI

SCI (N=103) SSCI (N=35)

Method 19%

System 20% Effects

33%

Audience 25% Audience

24%

Content 19%

% o

f Stu

dies

8

9

Common Themes Emerged

7. Opinion Leaders 8. Privacy Concerns 9. Recommender System 10. Social Capital, Social

Movements, Collective Actions

11. Spammer/Zombie Detection

12. User Behavior, Usage Pattern

1. Classification/Clustering 2. Cross-Site Comparison 3. Community /Group

Detection 4. Hot/Crisis Events 5. Information Diffusion

/Propagation 6. Keywords /Opinion

/Sentiment Extraction

Common Themes by 5Ws

Communicator Content Audience Channel Effects

Opinion Leaders Topic Classification Information Diffusion

Spammer /Zombie

Detection

Keywords /Opinion

/Sentiment Extraction

User Behavior;

Usage Pattern

Cross-Site Comparison

Social Capital; Collective

Actions

Recommender System

Hot/Crisis Events

Community /Group

Detection

Privacy Concerns

10 SCI; SSCI; Both

11

1. WHO (Communicator)

What we know: Media professionals Elite opinion leaders Government mouthpieces Rumor mills (“50-cent

Party”, “Water Army”, Spammers, etc.)

What we don’t know: Grassroots opinion

leaders E-marketers Foreign

individuals/entities

Communicators on Early Weibo (N=10)

0

5

10

15

20

Media (N=2) Ordinary Users (N=4) Grassroots OpinionLeaders (N=2)

Celebrities & VIPs(N=2)

N o

f Pos

ts/D

ay

Source: Xie (2010, in Chinese). Structure and mechanisms of interaction on Sina Weibo. 12

Government Presence on Social Media

Unique features: o Centralized review process o Widespread “Water Army” o Adaption of international practices

Fundamental problem: o Open-networked society vs. closed-centralized government

13 Source: Zhang (2013). Social media in Chinese government.

14

Growth of Weibo Accounts by Police Bureaus

Source: Ma (2013). The diffusion of government microblogging.

15

Predictors of Use of Weibo by Police Bureaus

Significant predictors: Government size Population size Adoption of neighbors Adoption of upper-level

Non-significant predictors: Revenue GRP Openness Safety level E-government efforts User size

Source: Ma (2013). The diffusion of government microblogging.

16

2. Says WHAT (Content)

What we know: Multiple duplicates Contentious Censorship

What we don’t know: Benchmark categorization Non-political content

(e.g., entertainment) Forced vs. self-censorship

Categories of Early Weibo Posts (N=438)

News 31%

Diaries 28%

Jokes 19%

Quotes 10%

Promotions 6%

Others 4%

Comments 3%

17 Source: Xie (2010, in Chinese). Structure and mechanisms of interaction on Sina Weibo.

Who Say What on Early Weibo

52% 36%

33%

71%

42%

36%

22%

19% 8%

0%

25%

50%

75%

100%

Media Ordinary Users GrassrootsOpinion Leaders

Celebrities &VIPs

CommentsOthersPromotionsQuotesJokesDiariesNews

18 Source: Xie (2010, in Chinese). Structure and mechanisms of interaction on Sina Weibo.

Sina Weibo vs. Twitter on Content

Weibo Twitter

Trending Topics Trivial jokes News events around the world

Trend Setters Unverified accounts Leading news media

Trendy Topics Retweeted 62% 31%

Presentation content Images, videos, links Text

Source: Yu et al. (2011). What trends in Chinese social media. 19

What Might Cause the Differences?

The higher rate of retweeting is likely to be created by spammers, i.e., the accounts that were set up to retweet (promote) targeted users, which also promote the relevant topics to be trending.

About 1% of the users are suspected spammers, who contribute 1/3 of the total posts or half of the retweeted posts.

If the spammers are removed, the evolution patterns of trending topics on Weibo becomes similar to that on Twitter.

20 Source: Yu et al. (2012). Artificial Inflation.

Detecting Spammers by User-Retweet Ratio

0%

25%

50%

75%

100%

35 25 15 10 9 8 7 6 5 4 3 2 1

% o

f Sam

ple

N of Users Retweeted by the Account

Inactive AccountActive Account

21 Source: Yu et al. (2012). Artificial Inflation.

Chinese Censorship

King et al. (2013): The largest selective suppression of human expression

in history: o implemented manually (within a few hours of posting) o by 200,000 workers o located in government and inside social media firms

A huge censorship organization: o (obviously) designed to suppress information o (paradoxically) very revealing about the goals, intentions,

and actions of the Chinese leadership

22 Source: King et al. (2013). Reverse engineering Chinese censorship.

23

What’s Censored

Source: King et al. (2013).

24

On the Ongoing Events in Hong Kong

25

3. Through WHICH Channel

What we know: Weibo concentrated Compared with Twitter Compared with old media

What we don’t know: Other social media, e.g.,

o IM (WeChat) o BBS (Tianya) o SNS (QQ) o Blogs, dating, gaming, etc.

Competition/cooperation among social media

26

Blogs vs. Microblogs

Method: 300 Sina blog posts on

traveling to Hong Kong 300 Sina Weibo posts on

the same topic Manual content analysis

of the frequency, direction, and intensity of keywords, themes, visuals, etc.

Findings: Blog posts largely to

recount past experience Weibo posts mainly to

express desire to visit in advance

No formal comparison made as the two samples treated as replicates rather than contrasts

Source: Tse & Zhang (2012). Analysis of Blogs and Microblogs

27

Competitor or Partners

PPM Model: Method: Drew a random sample of

Sina users Mined their blog/Weibo

posts to measure behavior Invited them to answer an

online questionnaire to measure motivation

Found most users to co-use both simultaneously

Source: Lu (2013). The sustainability of UGC.

Switching from Blogging to

Microblogging?

Push Forces (satisfaction)

Pull Forces (alternative

attraction, etc.)

Mooring Forces (perceived

popularity, etc.)

28

4. To WHOM (Audience)

What we know: Massive size Skewed participation Polarized views

What we don’t know: User authenticity Passive audience Inactive audience

Massive Size of Social Media Users

0

250

500

750

1000

Social media users(CNNIC)

Sina Weibo WeChat QQ

Mill

ion

Sources: CNNIC survey report; Sina/Tencent annual reports

Registered Monthly Active Daily Active

29

Changes in Popularity of Chinese Social Media

0%

25%

50%

75%

100%

Dec-08 Dec-09 Dec-10 Dec-11 Dec-12 Dec-13 Jun-14

% o

f Web

Use

rs S

urve

yed

Source: CNNIC Annual Survey Reports (www.cnnic.cn)

IMBlogMicroblogSNSBBS

30

31

Political Preference of Dominant Groups

Source: Wu (in press). Ideological polarization.

32

A 3-step Flow Model of Information Diffusion

Source: Chen (2013). Opinion Leaders: The Driving Force of Political Discussion in Social Media

Distinctions between Communicator and Audience

Participation Duration Content Creation Content Consumption

Continuous I. Active Communicator III. Active Audience

Interrupted II. Inactive Communicator (traceable)

IV. Inactive Audience (completely ignored)

33 Source: Zhu et al. (2014, in Chinese). Computational social science.

34

5. With WHAT Effects

What we know: Registration = Exposure Exposure = Conversion Expression = Participation

What we don’t know: Definition of effects Links to content Links to offline life Effects on individuals vs.

society

35

Expression vs. Participation

Arguments: Weibo helps rural migration workers to mobilize to improve their

life in cities; release emotional

dissatisfactions with reality;

influenced by journalists, scholars, and officials (unreciprocal weak ties?).

Method: Searched “new generation

of rural workers” (新生代农民工) to obtain 4,000+ posts;

Content analyzed the themes of the posts;

Results considered to support/derive the arguments.

Source: Zhang (2013). Social inclusion or exclusion?

Evidence for Arguments 1-2 (?)

0%

10%

20%

30%

40%

50%

Information Emotions Actions Others

Non-Emotion Emotion

Source: Zhang (2013). Social inclusion or exclusion? 36

37

Agenda-setting Effects based on N of Posts

Hypotheses: “Safety valve” effects:

bloggers express discontent on issues initiated by the media;

“Pressure cooker” effects: bloggers lead issues to increase social tensions.

Method: Randomly searched

through Google 2,000+ blog posts and 4,000+ newspaper stories;

Computerized content analysis of issue themes;

Time series analysis to determine causal directions between the two agendas.

Source: Hassid (2012). Safety valve or pressure cooker.

38

Safety Valve or Pressure Cooker?

Findings: Rather than simply being a ‘‘safety valve’’ that reduces bloggers’ anger over political or social problems, or a pressure cooker that increases these tensions, the results of this study suggested a more nuanced perspective in which the Chinese blogosphere can act as either, depending on the topic.

Safety value issues: o Energy; o Politics; o Corruption and illegality; o Sino-Japan relations

Pressure cooker issues: o Internet; o Religion; o Rural; o Arts; o Disasters.

Source: Hassid (2012). Safety valve or pressure cooker.

39

Summary of Empirical Evidence

What we know enough: Massive size of users Content censored or

manipulated Active and influential

opinion leaders

What we know little: Does the size matter? Is the censorship

effective? Causality: Does social

media bring social changes?

Discussion 1: State of the Art

Research on Chinese social media is a fast growing and increasingly popular field; prevailed by optimism by liberalists, marketers,

investors, and even the new generation of Chinese leadership;

commonly examined with case studies, focusing on crisis events, controversial issues, or active groups;

shortage of credible evidence for causal connections.

40

Discussion 2: Common and Old Problems

The conceptual enthusiasm and empirical shortfall are not specific to research on Chinese social media;

The same patterns occur in research on social media in general, probably at a slightly milder scale;

These are also longstanding issues in research on traditional media worldwide;

One common root of the above domains lies in the difficulty in getting necessary and quality data for the relevant research questions.

41

Price-Earning Ratio of Social Media

-50

0

50

100

150

Sina Tencent Renren Weibo Facebook Google Twitter

42

Industry Average Price-Earning Ratio

0

10

20

30

40

50

60

70

80

90

100

Internet Telecomm Oil Banking Real Estate

Pric

e/Ea

rnin

g

ChinaU.S.

43

Discussion 3: Call for Smarter Research Design

Need thoughtful and creative research designs, e.g., o online experiments to ensure causal inference o individual-level analysis to ensure ecological validity o cross-national comparisons to enhance generalizability

44

45

Discussion 4: Opportunities and Challenges

Opportunities: Unique characteristics of

the ecosystem behind GFW

Active and sizable scholars in China for possible collaboration

Increasing availability of open/affordable data for academic research

Challenges: How to engage in and

contribute to mainstream research?

How to find a dream collaborator with complementarity?

How to evaluate data quality and interpret peculiar or counter-intuitive results?

Cited Works 1. Chen, C. (2013). Opinion leaders: The driving force of political discussion in social media. Master’s

thesis, Middle Tennessee State University. 2. Hassid, J. (2012). Safety valve or pressure cooker? Blogs in Chinese political life. Journal of

Communication, 62, 212-230. 3. King, G., Pan, J., & Roberts, M. E. (2013). Reverse engineering Chinese censorship through

randomized experimentation and participant observation. Presented at the annual meeting of American Political Science Association, Chicago.

4. Lu, H. (2013). The sustainability of UGC: Understanding continuance and switching produsage Behavior. Ph.D. Dissertation, City University of Hong Kong.

5. Tse, T. S. M., & Zhang, E. Y. L. (2013) Analysis of blogs and microblogs: A case study of Chinese bloggers sharing their Hong Kong travel experiences. Asia Pacific Journal of Tourism Research, 18(4), 314-329.

6. Wu, A.X. (In press). Ideological polarization over a China-as-superpower mindset: An exploratory charting of belief systems among Chinese Internet users, 2008-2011. International Journal of Communication.

7. Xie, Y. H. (2010, in Chinese). Structure and mechanism of interactions on Sina Weibo: An empirical study. Journalism & Communication Research (Xinwen yu Chuanbo Yanjiu), 4, 60-69.

8. Yang, G. B. (2010). The power of the Internet in China: Citizen activism online. Columbia University Press.

46

Cited Works (cont’d) 9. Yin, L. G. (2010, in Chinese). Framing the Deng Yujiao case: How online opinion impacts offline

media reports. International Journalism (Guoji Xinwenjia), 2010(9), 25-31.

10. Yu, L., Asur, S., & Huberman, B. A. (2011). What trends in Chinese social media. Proceedings of The fifth ACM workshop on Social Network Mining and Analysis. ACM.

11. Yu, L. L., Asur, S., & Huberman, B. A. (2012). Artificial inflation: The real story of trends and trend-setters in Sina Weibo. ASE/IEEE International Conference on Social Computing.

12. Zhang, P. Y. (2013). Social inclusion or exclusion? When Weibo (Microblogging) meets the “New Generation” of rural migrant workers. Library Trends, 62(1), 63-80.

13. Zhang, L. (In press). Social media in Chinese government: Drivers, challenges and capabilities. Government Information Quarterly.

14. Zhou, Y. Q., & Moy, P. (2007). Parsing framing processes: The interplay between online public opinion and media coverage. Journal of Communication, 57, 79-98.

15. Zhu, J. J. H., Peng, T. Q., Liang, H., Wang, C. J., Qin, J., & Chen, H. X. (2014, in Chinese). Computational social science in communication research. e-Science Technology and Applications (Keyan Xinxihua Jishu he Yingyong), 5(2), 3-13.

47

THANK YOU & CONTACT US @

mailto: j.zhu@cityu.edu.hk http://weblab.com.cityu.edu.hk

Diverse Research Outlets Official statistical reports:

o e.g., CNNIC biannual surveys Listed-firm financial reports:

o e.g., Sina, Tencent, etc. Web traffic monitors:

o e.g., Alexa, Dratio (wrating.com), etc. Marketing/consulting research reports:

o e.g., McKinsey, iResearch, etc. Academic research papers:

o English language; Chinese language

49

50

Accurate Data vs. Insightful Analysis Ac

cura

cy

Insightfulness

Revenue Model of Top 4 Social Media, 2011

0

50

100

150

200

250

300

Tencent Easenet Sina Sohu

OthersGamesAd

Sources: the 2011 financial reports of Tencent, Easenent, Sina, and Sohu 51

52

CNNIC: The 34th Internet User Survey Report (July 2014)

By June 2014:

Total users = 632M

Penetration rate = 47%

Mobile users = 83%

0%

25%

50%

75%

100%

IM Blog

Microblog

SNS

BBS

Users of Social Media

53

McKinsey Global Institute Report (July 2014) China’s digital transformation: The Internet’s impact on productivity and growth

Acknowledgement: “... This independent MGI initiative drew on data and expertise from the Alibaba Group Research Center and Baidu Development Research Center. Part of our analysis was made possible through this collaboration...”

Cover-page data:

54

Power of Social Media “Readers seeking predictions about how all of these changes will affect political outcomes in China over the next few decades are likely to be frustrated. This is not intended as a predictive book, and Mr. Yang wisely makes no attempt to forecast future events based on current trends. But for an in-depth understanding of how life has changed for China’s 300 million plus Internet users, and how these personal transformations have in turn affected contemporary Chinese society and culture, Mr. Yang’s work is essential reading.”

Growth of SSCI Publications on Social Media

0

10

20

30

40

50

60

70

80

90

100

0

200

400

600

800

1000

1200

2006 2007 2008 2009 2010 2011 2012 2013

N o

f Art

icle

s on

PRC

per Y

ear

N o

f Tot

al A

rtic

les p

er Y

ear

TotalChina

55

Disciplines Involved in Social Media Research

0%

25%

50%

75%

100%

Total (N=4600+) China (N=280)

OtherHumanitiesCommunicationHealth & EnvirnonmentEducationSocial SciencesBusiness & EconomicsScience & TechnologyComputer & Information

56

57

WB

58

59

60

61

62

63

64

Studies on Chinese Social Media at WISA2014

Jing Luo and Lizhen Xu: Identification of Microblog Opinion Leader Based on User Feature and Interaction Network

Chong Kuang, Zhiyuan Liu, Maosong Sun, Feng Yu and Pengfei Ma: Quantifying Chinese Happiness via Large-Scale Microblogging Data

The 5W Model for Assessment

Harold Lasswell (1948): Communication is a process of

who says what to whom through which channel with

what effects.

66

WHAT Effects

WHOM Audience

WHICH Channel

WHAT Content

WHO Communi

-cator

Communicator Content Channel Audience Effects